First, a reframe that makes this whole job easier: you're not trying to lower a number, you're trying to fix the matches behind it. Markers read the match list, not the percentage — a well-explained 22% beats a gamed 12% every time (we've covered how scores are actually judged). So here's the pass to run on your own report, match by match, plus the "tricks" that now hurt more than they help.
Step 1: sort your matches into four buckets
Open the report and classify every significant match. Each bucket has a different fix:
- Quotes without quotation marks or citations. The dangerous bucket. Add the quotation marks and the citation — the match stays visible, but it converts from "problem" to "scholarship." Never fix a quote by lightly rewording it; that turns honest quoting into patchwriting.
- Paraphrases that track the source too closely. Same sentence skeleton, swapped synonyms. Rewrite these from understanding: close the source, explain the idea aloud in your own words, write that down, cite it. Structure changes; the match dissolves for real.
- Boilerplate. Assignment briefs, cover sheets, standard methods phrasing, common academic stock phrases. Delete what shouldn't be in the document (the brief, the template text); leave what must be there and stop worrying about it — markers recognize boilerplate instantly.
- Your bibliography and properly quoted text. Often excluded by the instructor's filters anyway. Not your problem; don't mangle your references to chase a number.
Step 2: fix the writing habits that create matches
Most inflated similarity comes from one drafting habit: writing with the source open and "translating" it sentence by sentence. The reliable alternative is note-based drafting — read, take short notes in your own phrasing, close the tab, write from the notes. It's slower per paragraph and dramatically cleaner per document, and it's also the habit that protects you from accidental plagiarism in the first place.
What not to do: the tricks that backfire
- Hidden text and character swaps. White-text padding, Cyrillic look-alike letters, invisible characters — Turnitin explicitly detects text manipulation now and surfaces it to instructors as a flag. You'd be converting "sloppy citation" into "deliberate deception," which is a far worse conversation.
- Synonym-spinners and "similarity reducers." They produce patchwriting with an AI accent — often still matched, newly at risk on the AI report, and always worse prose.
- Image-of-text and PDF tricks. Unreadable text means an unprocessable submission, which fails in its own way — and markers have seen every version of this.
Step 3: re-check the final version
Edits move matches in both directions — a rewritten paragraph can surface a new match you didn't know was there. Whatever route you use to check (a university draft slot or a real Turnitin check in no-repository mode), the version worth checking is the one you'll actually submit. Then read the new match list with the four buckets above, fix what's real, and submit knowing the report holds up — which was always the actual goal hiding behind "lower my score."